Global
outsourcing
or
foreign
direct
investment:
Why
apple
chose
outsourcing
for
the
iPod
Chu-Ping
Lo
*
DepartmentofAgriculturalEconomics,NationalTaiwanUniversity,Taipei,Taiwan
1. Introduction
Helpmanet al.(2004)arguedthatthemostproductivefirms preferforeigndirectinvestment(FDI)toarms-lengthtradewhen facedwithaproximity-concentrationtrade-off.Throughempirical studies,HeadandRies(2003)examinedthecaseofJapan,Girmaet al.(2005)examinedthecaseoftheUnitedKingdom,andWagner (2007)examinedthecaseofGermany,withallinargumentsthat themostproductivefirmspreferFDIoverexports.Thisassertion alsogainstheoreticalsupportfromAntra`sandHelpman (2004), whoappliedanincomplete contractmodeltodemonstratethat mostproductivefirmspreferFDIandless-productivefirmsprefer internationaloutsourcing.WhileHelpman et al.(2004)assume positivetransportationcostsandfirmsthereforefacethechoiceof whetherornottoexport,transportationcostsareassumedtobe absentintheAntra`s–Helpman’s(2004)model.Instead,Antra`sand Helpman (2004) apply an incomplete contract distortion to addressthedifferencebetweenFDIandinternationaloutsourcing strategies.Inall,bothapproachesassertthatthemostproductive firmsprefertoFDI.
However,someoftheworld’smostproductivefirms,especially intheelectronicsindustry(e.g.,Apple,Microsoft,Nokia, Hewlett-Packard)1 carry out international outsourcing intensively and
operate almost no FDI activities in launching their innovative products (e.g., iPod, game consoles, cell phones and notebook computers,respectively).Forexample,inOctober2001,Applefirst marketedtheiPod,whichwasa‘‘disruptiveinnovation’’withasetof innovativefunctions(ClaytonandRaynor,2003).Beingthemost revolutionaryportablemediaplayerinhistory,andinthesixyears afteritslaunch,Applesoldover100millioniPodsworldwide(see Apple’s‘‘TheBestGoingOn’’conference).ThesuccessofiPodover competingmusicplayerscamefromradicalinnovationsinits user-friendlyinterfaceandsmoothintegrationwithMacandWindows throughwhichuserscouldeasilypurchaseanddownloadmusic/ videocontent.Appledesignedthemajorityofthesystem architec-ture of the iPod in-house but simultaneously outsourced the remaining four hundred-plus intermediate components to both domesticandinternationalcontractors.2Appleevencontractedout
ARTICLE INFO Articlehistory: Received5June2010
Receivedinrevisedform22February2011 Accepted30June2011 JELclassification: F02 F14 Keywords: Electronicsindustry Internationaloutsourcing Incompletecontract FDI R&D ABSTRACT
Asimplemodelispresented,whereafirm’sproductivityisendogenizedbyitsR&Dinvestment.Itshows thatthemostproductivefirmsmaypreferinternationaloutsourcingtoforeigndirectinvestment(FDI)in industrieswithahighinnovationshare.Thehighinnovationsharemotivatesthefirmstoeconomizeon organizationalcostinordertosaveresourcesforR&Dinvestment,makingoutsourcingpreferabletoFDI becausetheformerincursasmaller organizationalcost.Thismodel helpsexplainwhyAppleInc., belongingto theelectronics industry,whichhasaparticularlyhighinnovationshare,launchedits innovativeiPodthroughinternationaloutsourcinginsteadofFDI.
ß2011ElsevierB.V.Allrightsreserved.
*Tel.:+886233662653. E-mailaddress:cplo@ntu.edu.tw.
1
BusinessWeekmagazineandtheBostonConsultingGroup’srankingof‘‘The World’sMostInnovativeCompanies’’in2006gaveAppleComputerInc.thetop rank,Microsoftthefifth,Nokiatheeighth,Sonythethirteenth,andHewlettPackard theforty-second.
2Afewhighcostcomponentsweremadebyothercompanies,suchasJapan’s
Toshiba(harddriveforonboardstorage),Korea’sSamsung(flashmemoryfor temporary storage), and U.S.-basedBroadcom (integrated videoprocessor for mediaplayback)(Lindenet al.,2007).
ContentslistsavailableatScienceDirect
Japan
and
the
World
Economy
j ou rna l hom e pa ge : w w w. e l s e v i e r. co m/ l o ca t e / j we
0922-1425/$–seefrontmatterß2011ElsevierB.V.Allrightsreserved. doi:10.1016/j.japwor.2011.06.002
itscoresoftware;forexample,theuserinterfacewasdesignedwith the help of Pixo, and the reference platform was designed by PortalPlayer(Kahney,2006).Ofthe$299retailpriceforthe30GB videoiPod,Appleaccountedfora25percentshareofthevaluefor thecoresoftwaredesign,while36percentisattributedtocontracted manufacturersin the low-wageAsia-Pacific region that provide generic components and direct labor (Linden et al., 2007).3
Examplesofthegenericcomponentsanddirectlaborincludethe metal casing, provided by Taiwan’s Foxconn, and assembly, completedinChinesefactoriesownedbyTaiwanesefirmssuchas InventeeAppliancesandFoxconn(Einhorn,2007).Thepatternof globaloutsourcingalsooccurswithotherelectronicsproducts,such asHewlett-Packard’snotebookcomputers,with39percentofthe valueofanotebookcomputerattributabletogenericinputsand directlabor.
In contrast to the outsourcing literature (e.g., Antra`s and Helpman,2004;Antra`s,2005),whichhasarguedthatthemost productive firms prefer FDI to international outsourcing especially when the product is at the innovative stage, both AppleandHP haveworkedclosely withoutsidepartnersfrom thebeginningsofdesign. Thisapproachhasbeendramatically undertaken by the lead firms in the electronics industry (Sturgeon, 2002).
Whyistheninternationaloutsourcingprevailingforthelead firms(i.e.,mostproductivefirms)intheelectronicsindustries? Theliteratureofglobalvaluechainprovidesanexplanation:itis the advances in information technology (IT) enabling the electronicsindustry to more easily standardize the interfaces betweencomponents,suchthatthecomponentmodularization hasmadeinternationaloutsourcingmoreattractivebyallowing multiplefirmstosharethesamegenericcomponentsatalower cost due to economies of scale (Sturgeon and Lee, 2001; Sturgeon, 2002).4 However, a shortcoming of the product
modularization hypothesis is that the lead firms in the electronicsindustry(e.g.,ApplecomputerandHewlett-Packard) have contracted out not only generic but also non-generic components.
Another key factor facilitating the growth of international outsourcing in the product modularization hypothesis is the lead firms’ desire to reduce fixed organization investment, especiallyinmanufacturingfacilities(SturgeonandLee,2001).I borrowthisconceptinthispaper,andarguethattheprevailing dependenceonglobaloutsourcingintheelectronicsindustryis largelyduetoNorthern firms’leveraging themassproduction capacity of those in the South. For example, Flextronics,5 a
contractelectronicsmakerheadquarteredinSingapore,worked withMicrosoftintheinitialstageofdeveloping itsinnovative game console, Xbox, by building customized (as opposed to generic)mechanical parts andperforming system tests. More importantly was Flextronics’ capacity to ramp up production fromzeroto100,000consolesperweekinlessthanfiveweeks. Production capacity was important because ‘‘if the system [Xbox] could not be built cost effectively and ramp up in volumes, it would have little chance against the more establishedsystemsonthemarket’’(Carbone,2002).Therefore, theinnovativeMicrosoftfocusedonR&D activitiesin orderto develop a game system that was technologically superior to rivalsystems(e.g.,Sony’sPlayStation),andthenoutsourcedthe
low-techcomponentstothesecontractfirmsthatspecializedin massproduction,therebyavoidingthesubstantial organization-al costs of establishing a production capacityabroad through FDI.6Withthesespecializedcontractmanufacturersaroundthe
world, theworld’s mostproductive andinnovativefirms, well known for their R&D capabilities, may prefer international outsourcingtoFDI as a waytocut costs andimprovegeneral productivitywhenlaunchingtheirinnovativeproductsbecause FDI incurs a larger fixed organization cost than international outsourcing.
The purpose ofthis studyis todetermine why the world’s mostproductiveandinnovativefirmsmaypreferinternational outsourcing to FDI.I incorporate Griliches’ (1986) production function into Antra`s–Helpman’s(2004) modelto argue thata Northern firm determines its global organizational structure basednot only ona trade-offbetween low-waged production costsandincompletecontractdistortion,butalsoonatrade-off between R&Dinvestment andorganizational cost.Iarguethat an increase in the innovation share augments the trade-off between R&D investment andorganizational cost, whichmay takeprecedentoverthetrade-offbetweenproductioncostsand incomplete-contract distortion when the innovation share is sufficientlylarge.Asithappens,inindustrieswithasufficiently high innovation share, the most productive firms tend to carry out international outsourcing activities rather than FDI becausethehighinnovation shareenhances theadvantagesof R&D investment. On the other hand, in industries with a relatively low innovation share, the most productive firms prefer FDI tooutsourcing becauseoutsourcing incurs a larger incomplete-contractdistortion.
Theremainderofthispaperisorganizedasfollows.InSection2, I revisit Griliches’ (1986) results by applying data from the Industrial R&D Investment Scoreboard reported by Eurostat to determinehowtheinnovationsharevariesacrossindustries.In Section3,Iadd toAntra`s–Helpman’s(2004)modelofGriliches’ (1986) production function, in which firms’ productivity is augmentedbyinnovation,andIarguethatthisdifferenceinthe innovationsharereshapesafirm’sglobalorganizationalform.In particular, the ordering in Antra`s–Helpman’s (2004) model is reversedwhenanindustryhasasufficienthighinnovationshares. Section4concludes.
2. R&Dinvestmentandproductivity
Itis welldocumentedthat researchand development hasa positive effect on productivity (e.g., Leonard, 1971; Mansfield, 1980; Griliches, 1980, 1986; Lichtenberg and Siegel, 1991; Meddaet al.,2003).7Theoretically,Griliches(1986) definedan augmentedCobb Douglasproductionfunction withaninput of R&DinvestmentandillustratedthatR&Dcontributesto produc-tivitygrowthinU.S.manufacturing.8ByusingtheNationalScience
Foundationdatasetonabout1000ofthelargestU.S. manufactur-ingfirmsfrom1957through1977,Griliches(1986)foundthatthe innovation share,measured by thecontributionof theR&D on
3
Distributionandretailsaccountedforanother25percent.InputsfromJapan accounted9percent,whileinputsfromTaiwanaccountedfor1percent.
4Nevertheless, theproductmodularizationisalso underwayinmanyother
sectors as well, such as apparel and footwear, toys,food processing, home furnishingsandlighting,brewing,andautomotiveparts,pharmaceutical produc-tion,butnotonlybeendramaticintheelectronicsindustry(Sturgeon,2002).
5
Foxconn also manufactures iPods for Apple, cell phones for Nokia, and computersforHewlettPackard,amongothers.
6InresponsetoMicrosoft’sgeneralproductioncostsavingfromoutsourcing,
SonyoutsourceditsPlayStationgameconsoletoFoxconn,aTaiwanesefirmwhose majorproductionfacilitiesarelocatedinChina.
7
Empirically,Griliches(1980)hasrevealedapositiveandrobustrelationship betweenR&DandtotalfactorproductivityfortheUnitedStates,asdidMeddaetal. (2003)forItaly.
8
TheproductionfunctioninGriliches’(1986)modelisgivenbyY=AIr
L1b
Cb(see alsoLichtenbergandSiegel,1991),whereLdenoteslaborandCdenotescapital. Here,Adenotesneutraltechnologystockthatisaccessibletoallproducers.R&D inputI=PisiKtiis a measure of thedistributed lageffect of past research
investmentsonproductivity,whereKtimeasurestherealgrossinvestmentin
researchinperiodi,andsiconnectsthelevelsofpastresearchtothecurrentstateof
productivity,ranges from0.09 to 0.17.9 In revisiting Griliches’
results, I use a more recent dataset from the Industrial R&D InvestmentScoreboardreportedbyEurostat,whichcollecteddata from more than 1,000 of the largest firms’ audited financial accountsandannualreportsfrom2002through2005.Thesefirms werepubliclylistedcompanieswithregisteredheadofficeslocated within the OECD member countries. However, the Scoreboard designatescompaniesinaccordancewiththeir sector classifica-tion,asdefinedbytheFinancialTimesStockExchangeIndex(ICB Classification).
AsshowninTable1,mycross-sectionalestimatedinnovation shareonOECDfirmsforallindustriesin2002–2005isabout0.12, whichissimilartoGriliches’sestimationthatofU.S.firmsin1966– 1977.10Furtherexaminationofindustry-specificinnovationshare
findsvariationamongindustries.Table1reportstheestimationof the innovation share for each industry, based on the ICB Classification codes for various industries. Industries with less than30observationsarenotpresented.AsshowninTable1,the estimated innovation share varies substantially among these industrygroups.Forexample,thecomputerhardwareindustryhas thehighestinnovationshareat0.46,andtheautomobilesandparts industryhasthelowestat0.05.Becauseofthevariationamong industries,Idividetheseindustriesintothreegroups:industries withanestimatedinnovationsharelargerthan0.30are catego-rizedas‘‘IndustrieswithHighInnovationShare’’;thosewithan estimated innovation share less than 0.15 are categorized as ‘‘Industries with Low Innovation Share’’; and the remaining industriesarecategorizedas‘‘IndustrieswithMediumInnovation Share.’’
Mostindustries have a low innovation share of somewhere around0.1 and belongto the‘‘Industries with LowInnovation
Share’’ group.Five industries have exceedingly large estimated innovation shares, ranging from0.30 to 0.46(0.4 on average). Amongthesefiveindustries,threehavethesamethree-digitICB code(957),andallhavethesametwo-digitICBcode(95).Inother words,thefiveindustriesbelongingtothe‘‘IndustrieswithHigh InnovationShare’’groupareallelectronicsandITindustries. 3. Themodel
TheseminalmodelofAntra`sandHelpman(2004)andAntra`s (2005)providethebasicstructureformyanalysis,wherelaborisa unique factor of production. Assume that, in the North-South world, there are J+1 industries, where J industries produce a continuumofdifferentiatedproducts,andtheremainingindustry producesahomogeneousgood.Thehomogeneousgoodisusedas a numeraire that can be freely produced and traded in both countriesinequilibrium.Consumershavesimilarpreferencesover alldifferentiated goods,andeach consumermaximizesa utility function: U¼x0þ 1 u XJ j¼1 Xu j; 0<u<1; (1)
wherex0representstheconsumptionofthehomogeneousgood,Xj isanindexofaggregateconsumptionofthedifferentiatedproducts in industry j, and u is a parameterrepresenting thedegree of substitution acrossdifferentiated-productindustries.The aggre-gate consumption in industry j, inturn,is a constant elasticity substitution(CES)utilityfunctionoftheconsumptionofdifferent varietiesoffinalgoodsxj(i):
Xj¼
Z
xjðiÞadi
1=a
; 0<
a
<1; (2)wheretherangeofvarietyiisendogenouslydetermined.Assume that
a
>u so that varieties are more substitutable within an industrythanacrossindustries.Thisleadstoaninversedemand functionforeachfinalgoodiaspjðiÞ¼Xuj axjðiÞa1; (3)
wherepj(i)isthepriceofthefinalgoodxj(i).
Table1
Cross-sectionalproductions:allindustries.Dependentvariable:lognetincomeofOECDfirms:2005.
Industries E.coefficient No.ofobservation R2
Allindustries 0.12(0.01) 1557 0.93
Industrieswithhighinnovationshareðˆr0:3Þ
Computerhardware(ICB9572) 0.46(0.08) 55 0.95
Software(ICB9537) 0.45(0.07) 122 0.86
Semiconductors(ICB9576) 0.43(0.05) 91 0.91
Telecommunicationequipment(ICB9578) 0.35(0.06) 61 0.94
Electroniccomponents(ICB2733) 0.30(0.07) 48 0.94
Industrieswithmediuminnovationshareð0:3> ˆr>0:15Þ
Householdgoods(ICB372) 0.27(0.07) 26 0.97
Healthcare(ICB453) 0.23(0.05) 76 0.97
Industrialmachinery(ICB2757) 0.16(0.04) 87 0.94
Industrieswithlowinnovationshareðˆr0:15Þ
Aerospaceanddefense(ICB271) 0.13(0.06) 40 0.98
Pharmaceuticals(ICB4577) 0.13(0.06) 85 0.91
Foodproducers(ICB357) 0.10(0.07) 39 0.91
Constructionandmaterials(ICB235) 0.08(0.05) 48 0.95
Chemicals(ICB135) 0.07(0.05) 109 0.90
Automobileandparts(ICB335) 0.05(0.05) 71 0.97
Netsalesexcludesalesofjointventuresandassociates.FollowingGriliches(1986),R&DstockistheaccumulatedsumofR&Dexpendituresfrom2002through2005basedon anassumptionof15percentdecliningbalancedepreciationperyear,wheretheR&Dexpenditureiscashinvestmentsfundedbycompaniesthemselves.Standarderrorsare showninparentheses.
Industrieswithfewerthan30samplefirmsarenotshown.Standarderrorsareshowninparentheses.Thethreeindustrygroupsarebasedontheirestimatedinnovation shareðˆrÞ.
9
PleasealsoseeEq. (4)foraclearerdefinitionoftheinnovationshare.To eliminatefixedeffects, Griliches(1986)analyzedthegrowth rateofadjusted productivityonthegrowthrateofR&Dinvestment,resultinginanestimationof 0.12,whichisveryclosetohisestimationinthecross-sectionallevelregression.
10Thebiotechnologyindustryisnotoriousforitsfledglingmarketsinwhichmany
firmshaveeithernegativenetincomesorenormouslylargeR&Dspending,soIrun tworegressions,onethatincludesthebiotechnologyindustryandanotherthat doesnot.Bothregressionshavealmostthesameresults.Incomparisontoother industries,onlythebiotechnologyindustryhasanegativeestimatedelasticity; however,itssmallsizedoesnotaffectourestimation.
In Antra`s–Helpman’s (2004) model, the unique final good producerofvarietyidrawsaparticularrealization
u
fromaknown distribution.11 Instead, in this current model, I argue that theproductivityisendogenouslydeterminedbyafirm’Researchand Development(R&D)decisionas:
u
¼AINðiÞrj; (4)where IN(i) denotes R&D investment made by the final good produceriintheNorthandAisanexogenousparameterdenoting theneutralproductivity.In(4),rjisanindustry-specificparameter denoting the extent to which skilled workers are efficiently employed, and 0<rj<1.12 It will be shown clearly in a later discussionthatrjincreasesanindustry’sR&Dintensity.Afirm’s productivityincreasesastheinvestmentinR&Dincreases,butin diminishingreturns.Obviously,incomparisontoindustrieswitha lower innovation share, a firm in an industry with a high innovationshare,ceteris paribus,is encouragedtoallocatemore resourcestoR&Dinvestmentawayfromorganizationinvestment inordertoraiseproductivityandrevenue.
Thefirmiinanindustryjisheterogeneousinproductivityand thenincorporates high-andlow-techinputsofHj(i) andMj(i) in production of the final good xj(i). The firm manufactures the high-tech inputsbutrelocatesthelow-techinputstotheSouthbyeitherFDIor internationaloutsourcing.Theproductionfunctionisthengivenby: xjðiÞ¼
u
HjðiÞ 1zj 1zj M jðiÞ zj zj ; 0<zj<1: (5)Thevariablezjrepresentsthematurityoftheindustryj,witha largerzimplyingthattheindustryisatamorematurestage.Let the wages of the North and South be denoted as wN and wS, respectively, and let wN>wS. Incorporating (4) into (5), the productionfunctionisrewrittenas
xj¼AINrj Hj 1zj 1zj M j zj zj : (6)
Becoming a Griliches-type (1986) production function, (6) presentsanincreasing-return-to-scalecharacteristic.For simplic-ity,theindustryindexjisdroppedfromallofthevariablesinthe followingdiscussion.
3.1. Divisionofsurplusandorganizationcosts
Tofocusonoffshoreproduction,Ionlydiscussthecaseswhere thelow-techinputsareproducedeitherbyFDIorbyinternational outsourcing.Asiswellknown,offshoreproductionislimitedby the incomplete nature of contracts governing international transactions.Duetothenon-verifiabilityoftherelevantstateof theworld,thepresenceofincompletecontractsleadstheparties involvedin offshoreproduction toacquire partialreturns,such that they may under-invest ex ante in relationship-specific investments. The threat of underinvestment motivates the Northern firm to purchase more residual rights from the subsidiarieslocatedintheSouthbyverticalintegration(FDI)in order to reduce the impact of potential underinvestment in comparingtointernationaloutsourcing.13
Therefore,atrade-offassociatedwiththeallocationofownership rightsarisesbetweenorganizationcostsandincomplete-contract
distortions. After a goodquality prototypeis provided, the two parties bargain over the surplus from the relationship-specific partnership.Ifsuchbargainingresultsinanagreementbetweenthe parties,from(3)and(5),thenthepotentialrevenuefromthesaleof thefinalgoodis:
RðiÞ¼Xua
u
a HðiÞ 1z að1zÞ MðiÞ z az : (7)However,ifthepartiesfailtoagree,theoutsideoptionofthe low-techinputsproducerwillbezerowhilethatofthehigh-techinputs firmswillvarywithdifferentorganizationalstrategies.Following Antra`sandHelpman(2004)andAntra`s(2005),theexpostsurplus shareisdependentonafirm’sorganizationalstructureas
b
K2(0,1), whereK2{V,O},whereVdenotesFDI,andOdenotesinternational outsourcing.Thatis,theNorthernfinal-goodproduceracquiresa fractionoftheexpostsurplus(i.e.,b
R(i))whiletheremainingsurplus isattributedtothelow-techinputprovider(i.e.,(1b
)R(i)).Antra`s andHelpman(2004)alsoassumedthatanFDIfirmacquiresmore leverage than one that outsources, resulting in the order ofb
V>b
O>1/2.14Notethatof the$299retailprice forthe30GB iPod,36percentofthevalueisattributedtogenericinputsanddirect labor provided in large part by firms in the Asia-Pacific region (Lindenet al.,2007).Similarly,fora$1400HPnotebookcomputer, about39percentofvalueisattributedtogenericinputsanddirect labor. Thus, atleast in the electronics industry, it is feasible to assumethattheb
Oisaroundone-halftotwo-thirds.Antra`sandHelpman(2004)alsoassumedthatFDIincurslarger fixedorganizationalcosts(investment),suchasfV>fO,wherefV and fO denote the fixed organizational costs for FDI and international outsourcing, respectively. In this paper,I use the sameorderingofdivisionofsurplusandorganizationcosts. 3.2. Equilibrium
Supposetherearenumerouslow-techinputsuppliersbidding forthelow-techinputcontractbypayingttotheNorthernfirms, wheretcanbeeitherpositiveornegative.Theincometransferwill cancel outin thejoint profitmaximization, so, forsimplicity, I neglectitinthefollowingdiscussion.AsinAntra`sandHelpman (2004)model,IassumeunitproductivityformanufacturingtheH andMinputsineithercountry.However,inthiscurrentmodel,the productivity is endogenously determined by a firm’s R&D investment,whichiscarriedoutbytheNorthworkers.
Withthe aboveset-up, a finalgood producermaximizes its profit:
max
IN;H
bK
RwNINwNHwNfK; (8)wherethefirm indexiisdroppedforsimplicity.Thefirst-order condition of (8) associated with the choices of H leads to
ab
K(1z)R=wNH,while that associated withthechoices ofIN leadstoa
rb
KR=wNIN.Theequilibriumshowsthatindustrieswith ahigherinnovationshare(i.e.,r)tendtohavehigherR&Dintensity (definedaswNIN/R=a
rb
K).Further,thelow-techinputsproducer maximizesitsprofitasmax
M ð1
bK
ÞRwSM: (9)Thefirst-orderconditionof(9)associatedwiththechoicesofM isgivenby
a
ð1bK
ÞzR¼wSM:11
AParetodistributionasinHelpmanet al.(2004).
12
Jones(1999)referredittoasthe‘‘steppingontoes’’effect(i.e.,thatduplication ofR&Deffortsismorelikelywhentherearetoomanypersonsengagedinit).The duplicationofactivitiesreducestheefficientuseofR&Dresources.
13
GrossmanandHart(1986)arguedthatcontractualrightsconsistofspecific rightsandresidualrights.Sinceitwouldbetoocostlytolistallthespecificrights overassetsinthecontract,soforeigninvestorsmightfinditadvantageousto purchasetheresidualrighttohavefullcontrolovertheasset.
14
GrossmanandHelpman(2005)assumed that,inthecaseofinternational outsourcing,Nashbargainingmakesthepartiesshareequallyinthesurplusthat accruesfromtheoutsourcingcontracts,thatis,bS=1/2.
Combining(8)and(9),theprofitoftheheadquartersfirmunder jointprofitmaximizationisthengivenby
pK
ðr;X;zÞ¼Xuau
aHað1zÞMazwNINwNHwSM
wNfK: (10)
Incorporating theabove first-order conditions into (10), we obtain
p
C K¼Xma=ð1aÞu
a =ð1aÞc
CKwNfKC; (11) wherec
Kðb
KÞ¼ð1a
rb
Ka
½b
Kð1zÞþzð1b
KÞÞ=½ð1=a
ÞðwN=bK
Þ1zðwS=1bK
Þza=1a (seeAppendix Aforthederivation).In(11),theprofitmaximizationisequivalenttothemaximizationofits slope,
c
K.Theoptimaldivisionofsharecanthenbederivedbytaking˙
cK
ðbK
Þ¼0;in particular,we obtainb
¼ðð1zþr=2Þð1a
zÞ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ½ð1zþr=2Þð1a
zÞ2ð12zþrÞð1a
zÞð1zÞ q Þ=ð12zþrÞ, which converges to the original Antra`s and Helpman (2004) expressionwhenr=0.As shown in Table 1, the majority of industries have an innovationshare of aroundr=0.1, but theelectronics industry groupalonehasahighinnovationshareofaboutr=0.4.Wecan plot
b
*(z)asafunctionofzforanindustryj,whichhasaspecificrj, bysimulationasshowninFig.1.Tosharpenthecontrast,Iintroduceonlytwodifferentindustry groupsinthesimulation:onewithahighinnovationshareof0.4 andtheotherwithalowinnovationshareof0.1.Thethincurvein Fig. 1 illustrates the case where r=0.1, and the curve in bold illustratesthecaseinwhichr=0.4.Specifically,the
b
*(z,r=0.4)in Fig.1denotestheelectronicsindustryandb
*(z,r=0.1)represents theotherindustries.Itiseasytofindoutb
*(0,r)=1/(1+r)andb
*(1,r)=0,suggestingthattheoptimalb
*fortheskill-intensive industryislowerifthecorrespondinginnovationsharerisgreater. Forcomparison,Ialsoillustratethedivisionofsurplusdetermined bybargainunderFDIandunderoutsourcingmodesinFig.1. 4. FDIvs.outsourcingOutsourcing strategy dominates FDI in the labor-intensive industries (i.e., z>0.5), as illustrated in Antra`s and Helpman (2004)model.Thus, inthis paper,Ifocus ontheskill-intensive industries(i.e.,z<0.5),inwhichourresultsdiffer.Tofocusonour analysis,Ialsoassumeinthismodelthatthewagegapbetweenthe North and the South is large enough so that the strategy of domestic production is preferred by all other organizational structures.
Inanindustrywithalowinnovationshare(asmallerr),
b
Vis closertotheoptimalb
*curvethanb
O,indicatingc
V>c
O.Then, wecanplottheprofitwithrespecttotheproductivityu
ai=1aforthelowinnovationshareindustriesasinFig.2.Asimpliedin(11), the slopes of the profit lines are crucially determined by
c
K, indicatingthattheFDImodeisamoreoptimalstrategythanthe outsourcingmodeforthemostproductivefirms.Thisresultisin linewithAntra`s–Helpman’s(2004)model.This current model deviates from Antra`s–Helpman’s (2004) modelwhentheinnovationshareissufficientlylarge.Fig.1shows that,thehighertheinnovationshareinanindustry,thecloserthe outsourcingmodeistotheoptimaldivisionofsurplus.Thatis,in anindustrywithahighinnovationshare(alargerr),
b
Oiscloserto theoptimalb
* curvethanb
V,indicatingc
V<c
O.Asshown in Fig.3,theslopeoftheprofitlineunderFDImodemaybesmaller than that of outsourcing mode, implying that the outsourcingFig.2.Equilibriuminindustrieswithalowerr. Fig.1.Distributionofb*ofelectronicsandtheotherindustries.
modemaybeamoreoptimalstrategythantheFDImodeandmore profitable.Asaresult,themostproductivefirmsinanindustry with a higher innovation share may prefer international out-sourcing to FDI. This is particularly true for the electronics industry,whichhasanespeciallylargeestimatedinnovationshare compared to that of other industries. I argue that the high innovation share motivates Northern firms to economize on organizational spending in order to save resources for R&D investment,therebymakingthefirmmoreproductive.Thus,the internationaloutsourcingmodeispreferredbecauseitrequiresa lowerorganizationalcostthanFDI.
As is well known, the opportunistic underinvestment may takeplacewhencontractsareincomplete(GrossmanandHart, 1986; Hart and Moore, 1988). The division of surplus
b
K is bargainedoverseekingtobalancethe underinvestmentofthe final-goodproducerandintermediate-inputsuppliers.Notethat eachparty’sseverityofunderinvestmentisinverselyrelatedto thefractionofthesurplusthatitappropriates. Iarguethatan increasein R&D investment, whichincreases productivityand revenuemoreinanindustrywithahighinnovationsharethan inanindustrywithalowone,enlargestheopportunitycostof underinvestmentfor both parties. As a result, the severity of underinvestmentis scaleddownwhentheinnovationshareof an industry is high. With less severity in opportunistic incentives, the profit-maximizing final good firms may adopt a structure of international outsourcing rather than FDI to economizeonorganizationalcosts.ThisargumentissomehowinlinewiththatofGlassandSaggi (2001), who presented a product cycle model to argue that shifting production to the South increases product market profits throughthe lower wagecost of production there,and frees up Northern resources for innovation. However, our approachinthispaperdiffersfromthatoftheirmodelinmany respects.Theirmodelisdrivenbydifferencesintechnologyand wageratesbetweentheNorthandSouth,whereasourmodelis driven by the differences in incomplete contracts and in organizationcosts betweenFDI and internationaloutsourcing. Theirmodelonlyaddressestheproblemofoffshoreproduction (international outsourcing only) or not while our analysis addressestheproblemofwhichmethodofoffshoreproduction (i.e., FDI or international outsourcing) may play a more important role in freeing up the resources of Northern firms inimprovingproductivity.
5. Conclusions
Thedifferenceininnovationshareamongindustriesimpacts thefirms’globalorganizationalform.Inanindustrywithahigh innovation share, the Northern firm tends to economize on organizationalinvestmentinordertoputmoreeffortsintoR&D investmentwhileR&Dinvestmentandorganizationalcostsare substitutive. That is, the firms in an industry with a larger innovation shareare encouraged todevote more resourcesto R&Dinvestment,ceterisparibus,inordertoincreaseproductivity andenlargerevenuebutwhilesimultaneouslychoosingalower cost of global organizational from in order to minimize the aggregate nonproduction costs. While an increase in revenue (duetoanincreaseinR&Dinvestment)enlargesanopportunity costofunderinvestmentforthetwopartiesinvolved,themost productive firms in those industries with a larger innovation sharetendtopreferinternationaloutsourcingtoFDIwhilethe incompletecontract distortion is restrained.This model helps explainwhytheworld’smostproductivefirms,suchasApple, Nokia, Microsoft, and HP (all in the electronics industry), often prefer international outsourcing to FDI when launching innovativeproducts.
Acknowledgment
ThefinancialsupportprovidedbytheNationalScienceCouncil (NSC96-2415-H-002-019)isgratefullyacknowledged.
AppendixA
From(8),we haveH¼ð
abK
ð1zÞ=wNÞRandIN¼ða
rbK
=wNÞR.From(9),wehaveM¼ð
a
ð1bK
Þz=wSÞR.PlugtheoptimalHandMinto(7),weobtain R¼Xua
u
aabK
wN R að1zÞa
ð1bK
Þ wN R az : (A1)Wecanrearrange(A1)as R¼Xua=1a
u
a=1aabK
wN að1zÞ=ð1aÞa
ð1bK
Þ wN az=ð1aÞ : (A2)Thejointprofitmaximizationin(10)isthengivenby
pK
ðr;X;zÞ¼Xuau
aHað1zÞMazwNINwNHwSM wNfK ¼Ra
rbK
RabK
ð1zÞRa
ð1bK
ÞzR wNfK ¼f1a
rbK
a
½bK
ð1zÞþð1bK
ÞzgR wNfK: (A3)Plug(A2)into(A3),weobtain
pK
¼Xma=ð1aÞu
a=ð1aÞcK
wNfK; (A4) wherec
Kðb
KÞ¼ 1a
rbK
a
½bK
ð1zÞþzð1bK
Þ ð1=a
ÞðwN=bK
Þ1zðwS=1bK
Þ z h ia=ð1aÞThe optimal division of share can then be derived by taking ˙
cK
ðbK
Þ¼0; It is equivalent to take derivative of f1a
rbK
a
½bK
ð1zÞþzð1bK
Þg½ðbK
Þzð1bK
Þ1za=ð1aÞwithrespecttob
K. Itistheneasytocalculatetheoptimaldivisionofshareb
*. ReferencesAntra`s,P.,Helpman,E.,2004.Globalsourcing.JournalofPoliticalEconomy112, 552–580.
Antra`s,P.,2005.Incompletecontractsandtheproductcycle.AmericanEconomic Review95,1054–1073.
Carbone,J.,2002.OutsourcingtheXbox.PurchasingMagazine.
Clayton, M.C., Raynor,M.E., 2003. Theinnovator’s solution.Harvard Business Review74,65–73.
Einhorn,B.,2007.Apple’sChinesesupplylines.BusinessWeekJanuary8. Girma,S.,Kneller,R.,Pisu,M.,2005.ExportsversusFDI:anempiricaltest.Reviewof
WorldEconomics2,193–218.
Glass,A.J.,Saggi,K.,2001.Innovationandwageeffectsofinternationaloutsourcing. EuropeanEconomicReview45,67–86.
Griliches,Z.,1980.Productivity.RandD,andbasicresearchatthefirmlevelinthe 1970s.AmericanEconomicReview70,343–348.
Griliches,Z.,1986.R&Dandtheproductivityslowdown.AmericanEconomic Review76(1),141–154.
Grossman,S.,Hart,O.,1986.TheCostsandBenefitsofOwnership:ATheoryof VerticalandLateralIntegration.JournalofPoliticalEconomy94(4),691–719. Grossman,S.,Helpman,E.,2005.OutsourcingInAGlobalEconomy.Reviewof
EconomicStudies72,135–159.
Hart,O.D.,Moore,J.,1988.IncompleteContractsandRenegotiation.Econometrica 56(4),755–785.
Head,K.,Ries,J.,2003.HeterogeneityandtheFDIversusexportdecisionofJapanese manufacturers.JournaloftheJapaneseandInternationalEconomies17,448–467.
Helpman,E.,Melitz,M.J.,Yeaple,S.R.,2004.ExportversusFDIwithheterogeneous firms.AmericanEconomicReview94,300–316.
Jones,C.I.,1999.IntroductiontoEconomicGrowth.W.W.Norton&Company. Kahney,L.,2006.StraightdopeontheiPod’sbirth.WiredNews10–17. Leonard,W.N.,1971.Researchanddevelopmentinindustrialgrowth.Journalof
PoliticalEconomy79,232–256.
Lichtenberg,F.R.,Siegel,D.,1991.TheimpactofR&Dinvestmenton productivi-ty: new evidence using linked R&D-LED data. Economic Inquiry 29, 203–229.
Linden, G.,Kraemer,K.L., Dedrick,J., 2007. WhoCaptures Value in a Global InnovationSystems:TheCaseofApple’siPod.PersonalComputingIndustry Center,IrvineCA,p.403.
Mansfield,E.,1980.Basicresearchandproductivityincreaseinmanufacturing. AmericanEconomicReview70,863–873.
Medda,G.,Piga, C.,Siegel,D.S., 2003. OntheRelationship betweenR&D and Productivity: ATreatment Effect Analysis. , RensselaerWorking Paper in Economics.
Sturgeon,T.,Lee,J.R.,2001.IndustryCo-evolutionandtheRiseofShared Supply-baseforElectronicsManufacturing.MITIndustrialPerformanceCenter, work-ingpaperno.01-002.
Sturgeon, T., 2002. Modularproductionnetworks: a newAmerican modelof industrialorganization.IndustrialandCorporateChange11,451–496. Wagner,J.,2007.Exportsandproductivity:asurveyoftheevidencefromfirm-level